def test_qcut_small():
nquantiles = [4, 5, 4, 2, 5, 4, 10, 3, 2, 5]
colnames = [
"bool", "one_group_odd", "one_group_even", "int_pos", "int_neg", "int",
"float", "inf_max", "inf_min", "inf"
]
DT = dt.Frame(
[[True, None, False, False, True, None], [None, 10, None, 10, 10, 10],
[None, 10, None, 10, 10, 10], [3, None, 4, 1, 5, 4],
[-5, -1, -1, -1, None, 0], [None, -5, -314, 0, 5, 314],
[None, 1.4, 4.1, 1.5, 5.9, 1.4], [math.inf, 1.4, 4.1, 1.5, 5.9, 1.4],
[-math.inf, 1.4, 4.1, 1.5, 5.9, 1.4],
[-math.inf, 1.4, 4.1, math.inf, 5.9, 1.4]],
names=colnames)
DT_ref = dt.Frame(
[[3, None, 0, 0, 3, None], [None, 2, None, 2, 2, 2],
[None, 1, None, 1, 1, 1], [0, None, 1, 0, 1, 1],
[0, 2, 2, 2, None, 4], [None, 0, 0, 1, 2, 3], [None, 0, 6, 3, 9, 0],
[2, 0, 1, 0, 2, 0], [0, 0, 1, 0, 1, 0], [0, 1, 2, 4, 3, 1]],
names=colnames,
stypes=[stype.int32] * DT.ncols)
DT_qcut = DT[:, qcut(f[:], nquantiles=nquantiles)]
DT_qcut_frame = DT[:, qcut(DT, nquantiles=nquantiles)]
assert_equals(DT_ref, DT_qcut)
assert_equals(DT_ref, DT_qcut_frame)
def test_qcut_str():
assert str(qcut(f.A)) == "FExpr<qcut(f.A)>"
assert str(qcut(f.A) + 1) == "FExpr<qcut(f.A) + 1>"
assert str(qcut(f.A + f.B)) == "FExpr<qcut(f.A + f.B)>"
assert str(qcut(f.B, nquantiles=3)) == "FExpr<qcut(f.B, nquantiles=3)>"
assert str(qcut(f[:2], nquantiles=[3, 4])) == \
"FExpr<qcut(f[:2], nquantiles=[3, 4])>"
def test_qcut_error_inconsistent_nquantiles():
msg = "When nquantiles is a list or a tuple, its length must be " \
"the same as the number of input columns, " \
"i.e. 2, instead got: 1"
DT = dt.Frame([[3, 1, 4], [1, 5, 9]])
with pytest.raises(ValueError, match=msg):
DT[:, qcut(f[:], nquantiles=[10])]
def test_qcut_random(pandas, seed):
random.seed(seed)
max_size = 20
max_value = 100
nrows = random.randint(1, max_size)
ncols = 4
stypes = (stype.bool8, stype.int32, stype.float64, stype.float64)
names = ("bool", "int", "float", "nafloat")
nquantiles = [random.randint(1, max_size) for _ in range(ncols)]
data = [[] for _ in range(ncols)]
for _ in range(nrows):
data[0].append(random.randint(0, 1) if random.random() > 0.1 else None)
data[1].append(
random.randint(-max_value, max_value
) if random.random() > 0.05 else None)
data[2].append(random.random() * 2 * max_value -
max_value if random.random() > 0.2 else None)
data[3].append(random.random() * 2 * max_value -
max_value if random.random() < 0.1 else None)
DT = dt.Frame(data, stypes=stypes, names=names)
DT_qcut = DT[:, qcut(f[:], nquantiles=nquantiles)]
DT_nunique = DT.nunique()
frame_integrity_check(DT_qcut)
assert DT_qcut.names == names
assert DT_qcut.stypes == tuple(stype.int32 for _ in range(ncols))
for j in range(ncols):
if DT_nunique[0, j] == 1:
c = int((nquantiles[j] - 1) / 2)
assert (DT_qcut[j].to_list() == [[
None if DT[i, j] is None else c for i in range(nrows)
]])
else:
if DT_qcut[j].countna1() == nrows:
assert DT_qcut[j].min1() == None
assert DT_qcut[j].max1() == None
else:
assert DT_qcut[j].min1() == 0
assert DT_qcut[j].max1() == nquantiles[j] - 1
def test_qcut_vs_pandas_random(pandas, seed):
random.seed(seed)
max_size = 20
max_value = 100
n = random.randint(1, max_size)
ncols = 2
nquantiles = [random.randint(1, max_size) for _ in range(ncols)]
data = [[] for _ in range(ncols)]
for _ in range(n):
data[0].append(random.randint(-max_value, max_value))
data[1].append(random.random() * 2 * max_value - max_value)
DT = dt.Frame(data, stypes=[stype.int32, stype.float64])
DT_qcut = DT[:, qcut(f[:], nquantiles=nquantiles)]
PD_qcut = [
pandas.qcut(data[i], nquantiles[i], labels=False) for i in range(ncols)
]
assert [list(PD_qcut[i]) for i in range(ncols)] == DT_qcut.to_list()
def test_qcut_error_negative_nquantiles_list():
msg = r"All elements in nquantiles must be positive, got nquantiles\[1\]: -1"
DT = dt.Frame([[3, 1, 4], [1, 5, 9]])
with pytest.raises(ValueError, match=msg):
DT[:, qcut(f[:], nquantiles=[10, -1])]
def test_qcut_error_negative_nquantiles():
msg = "Number of quantiles must be positive, instead got: -10"
DT = dt.Frame(range(10))
with pytest.raises(ValueError, match=msg):
DT[:, qcut(f[:], nquantiles=-10)]
def test_qcut_error_float_nquantiles():
msg = "Expected an integer, instead got <class 'float'>"
DT = dt.Frame(range(10))
with pytest.raises(TypeError, match=msg):
DT[:, qcut(f[:], nquantiles=1.5)]
def test_qcut_error_wrong_column_type_zero_rows():
DT = dt.Frame(obj=[] / dt.obj64)
msg = r"qcut\(\) cannot be applied to string or object columns, instead " \
"column 0 has an stype: obj64"
with pytest.raises(TypeError, match=msg):
DT[:, qcut(f[:])]
def test_qcut_error_noargs():
msg = r"Function datatable\.qcut\(\) requires exactly 1 positional " \
"argument, but none were given"
with pytest.raises(TypeError, match=msg):
qcut()
def test_qcut_one_row():
nquantiles = [1, 2, 3, 4]
DT = dt.Frame([[True], [404], [3.1415926], [None]])
DT_qcut = DT[:, qcut(f[:], nquantiles=nquantiles)]
assert DT_qcut.to_list() == [[0], [0], [1], [None]]
def test_qcut_expr_simple():
DT = dt.Frame([range(0, 30, 3), range(0, 20, 2)])
DT_qcut = DT[:, qcut(f[0] - f[1])]
assert_equals(dt.Frame(range(10)), DT_qcut)
def test_qcut_trivial():
DT = dt.Frame({"trivial": range(10)})
DT_qcut = DT[:, qcut(f[:])]
expr_qcut = qcut(DT)
assert isinstance(expr_qcut, FExpr)
assert_equals(DT, DT_qcut)
def test_qcut_zerorow_frame():
DT = dt.Frame([[], []])
DT_qcut = DT[:, qcut(f[:])]
expr_qcut = qcut(DT)
assert isinstance(expr_qcut, FExpr)
assert_equals(DT_qcut, dt.Frame([[] / dt.int32, [] / dt.int32]))
def test_qcut_empty_frame():
DT = dt.Frame()
expr_qcut = qcut(DT)
assert isinstance(expr_qcut, FExpr)
assert_equals(DT[:, f[:]], DT)
def test_qcut_error_groupby():
msg = r"qcut\(\) cannot be used in a groupby context"
DT = dt.Frame(range(10))
with pytest.raises(NotImplementedError, match=msg):
DT[:, qcut(f[0]), f[0]]
def test_qcut_error_wrong_column_types():
DT = dt.Frame([[0], [dt]])
msg = r"qcut\(\) cannot be applied to string or object columns, instead " \
"column 1 has an stype: obj64"
with pytest.raises(TypeError, match=msg):
DT[:, qcut(f[:])]
